Using a sample of nine massive compact galaxies at z~ 2.3 with rest-frame optical spectroscopy and comprehensive U → 8 μm photometry, we investigate how assumptions in spectral energy distribution (SED) modeling change the stellar mass estimates of these galaxies, and how this affects our interpretation of their size evolution. The SEDs are fitted to τ-models with a range of metallicities, dust laws, and different stellar population synthesis codes. These models indicate masses equal to, or slightly smaller than, our default masses. The maximum difference is 0.16 dex for each parameter considered, and only 0.18 dex for the most extreme combination of parameters. Two-component populations with a maximally old stellar population superposed with a young component provide reasonable fits to these SEDs using the models of Bruzual & Charlot; however, when using models with updated treatment of TP-AGB stars, the fits are poorer. The two-component models predict masses that are 0.08-0.22 dex larger than the τ-models. We also test the effect of a bottom-light initial mass function (IMF) and find that it would reduce the masses of these galaxies by 0.3 dex. Considering the range of allowable masses from the τ-models, two-component fits, and IMF, we conclude that on average these galaxies lie below the mass-size relation of galaxies in the local universe by a factor of 3-9, depending on the SED models used.